SPECTRAL CLUSTERING AND VISUALIZATION: A ... - Carl Meyer
SPECTRAL CLUSTERING AND VISUALIZATION: A ... - Carl Meyer
SPECTRAL CLUSTERING AND VISUALIZATION: A ... - Carl Meyer
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
BENSON-PUTNINS, BONFARDIN, MAGNONI, <strong>AND</strong> MARTIN<br />
The next step in the Fiedler Method is to take each subgraph and partition<br />
it using its own Fiedler vector. For our example, this second iteration works well<br />
on the right hand part of the graph (see Figure 3.4). For this small graph, two<br />
iterations are sufficient to provide well connected clusters, but as graphs become<br />
larger, more iterations become necessary. There are different ideas for when one<br />
should stop partitioning the subgraphs. We created an algorithm that creates a<br />
specified number of clusters, k.<br />
Fig. 3.4. Partition made by the second iteration of the Fiedler Method<br />
3.2. Limitations. The Fiedler Method has gained acceptance as a viable clustering<br />
technique for simple graphs. There are, however, some disadvantages to this<br />
method. First, the Fiedler Method is iterative, so if any questionable partitions are<br />
made, the mistake could be magnified through further iterations. Second, new eigendecompositions<br />
must be found at every iteration; this can be expensive for large data<br />
sets. Finally, this method was designed for undirected, weighted graphs. Unweighted<br />
graphs can be considered by assigning each edge to have weight 1. Directed graphs<br />
can be considered as well through utilization of additional eigenvectors of L, but we<br />
will not make use of these techniques.<br />
4. MinMaxCut Method. Like the Fiedler Method, the MinMaxCut Method is<br />
a spectral clustering method that partitions a graph. Spectral clustering methods use<br />
the spectral, or eigen, properties of a matrix to identify clusters. There are a number of<br />
spectral clustering methods, several of which are given in detail in [12]. Although the<br />
Fiedler Method and the MinMaxCut Method are both spectral clustering methods,<br />
the MinMaxCut Method can create more than two clusters simultaneously while the<br />
Fiedler Method creates two clusters with each iteration.<br />
4.1. Background. Before explaining the MinMaxCut Method, we describe a<br />
similar, more intuitive, algorithm: the Ratio Cut Method. The goal of this method is<br />
to partition an undirected, weighted graph into k clusters through the minimization<br />
of what is known as the ratio cut. Given a graph, such as a consensus matrix, broken<br />
into k clusters X 1 , X 2 , . . . , X k , the ratio cut is defined to be<br />
(4.1)<br />
k∑<br />
i=1<br />
w(X i , X i )<br />
|X i |<br />
where |X| is the number of vertices in X, X is the complement of X and, given two<br />
subgraphs X and Y , w(X, Y ) is the sum of the weights of edges between X and Y .<br />
6